๐Ÿ๏ธ Hwani Studio

Home โ€บ AI & IT news โ€บ 2026-11-20

AI news, November 20, 2026: Anthropic goes public, Stitch Loop, Genie 4 in the Gemini API

ยท Hwani Studio ยท AI, Anthropic, IPO, Google, Stitch Loop, Genie 4, Gemini API, Gemini, computer use, LangChain, agents

Five stories from Wednesday and Thursday, November 18โ€“19, 2026 (US time) for solo makers who build with AI. One is a market event for the company behind Claude, two are new Google agents that touch your code or your desktop, one is a world model you can now call from an API, and one is a governance layer for agents that call tools on your behalf. Figures are as the companies and outlets state them.

Anthropic prices its IPO at $83 a share and starts trading on Nasdaq as ANT

Anthropic priced its initial public offering at $83.00 per share for 980 million Class A shares, most of them sold by the company, with trading on the Nasdaq Global Select Market under the symbol "ANT" from November 19. The SEC pricing document lists an offering size of about $81.34 billion plus a concurrent $26.0 billion private placement at the same price. CNBC reports the stock opened at $111.50 and closed its first day at $97.17, about 17% above the IPO price, and AP calls it the biggest IPO on record.

Why it matters for makers: if Claude is part of your stack, the company behind it now reports to public shareholders, so pricing, plan and policy changes will come with more public disclosure than before.

Sources: Anthropic, SEC filing, CNBC, AP News

Google Labs opens a waitlist for Stitch Loop, a product agent that opens pull requests from live metrics

Stitch Loop is a long-running Google Labs experiment that connects to your GitHub repository and product context such as logs, user behavior and feedback, through more than 180 integrations, custom MCP servers or CLI uploads. Each loop suggests improvements grounded in that data, offers solutions with tradeoffs, writes the chosen change in an isolated cloud sandbox and opens a pull request, then keeps watching the metric after the goal is met. It ships with default loops for security, latency and accessibility, and Google stresses that nothing merges until you review the PR. The waitlist is open globally, with priority for eligible Google AI Ultra subscribers; Google has not published pricing.

Why it matters for makers: post-launch upkeep is where one-person products usually stall, and an agent that turns analytics into reviewable PRs targets exactly that gap, while keeping the merge decision with you.

Sources: Google, TestingCatalog

Genie 4 world generation is now available to developers in the Gemini API

A day after launching Genie 4 for AI Ultra subscribers, Google made it available through the Gemini API's Interactions API and in Google AI Studio. From a text prompt, a reference image or both, it generates continuous 720p worlds at 24 frames per second with synchronized native audio, which you can explore with keyboard controls and change with live text prompts, such as weather, lighting or objects. Google says every output carries an imperceptible SynthID watermark.

Why it matters for makers: interactive worlds from a prompt are a quick way to prototype game levels, spatial concepts or test environments for agents before you build anything by hand.

Source: Google

Gemini on Mac can now use your apps and websites for you

Google added computer use to the Gemini app for Mac, so it can navigate native apps and websites in a separate window and finish multi-step chores while you keep working. You choose which apps it may access and approve its actions, and TechCrunch notes you can pause or stop it at any point. Google's examples include filling out forms, repeating one task across a batch of files and organizing folders. It needs an Apple Silicon Mac on macOS Sequoia 15.0 or later and a Google AI Plus, Pro or Ultra plan, and TechCrunch reports it starts in English in the U.S.

Why it matters for makers: desktop agents are becoming a normal consumer feature, so admin pages and forms in your own tools should work with clear labels and predictable steps that an agent can follow.

Sources: Google, TechCrunch

LangChain launches LangSmith Tool Gateway to govern which tools agents can call

LangChain released LangSmith Tool Gateway in public beta, a managed layer that routes an agent's tool calls through a single URL and handles credentials, identity and access for each caller. It sits next to the existing LLM Gateway, which governs model calls with spend limits, fallbacks, rate limits and redaction. LangChain says each tool call is logged in the agent trace with whose credentials were used, which server answered, latency and outcome, aimed at agents that run on schedules or events without someone checking every step.

Why it matters for makers: once your agents run unattended, a per-tool permission list and an audit trail of every call are the basics to have in place, whether you use this product or build the same checks yourself.

Source: LangChain